Table 1.
Performance (in %) of MetaMap and Peregrine, with and without the NLP module, on the AZDC test set for exact and inexact matching of concept boundaries and identifiers
| System | Exact matching | Inexact matching | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Boundaries | Identifiers | Boundaries | Identifiers | |||||||||
| Precision | Recall | F-score | Precision | Recall | F-score | Precision | Recall | F-score | Precision | Recall | F-score | |
| MetaMap | 60.9 | 61.1 | 61.0 | 55.0 | 55.2 | 55.1 | 79.3 | 79.7 | 79.5 | 65.1 | 65.5 | 65.3 |
| MetaMap+NLP | 76.1 | 70.7 | 73.3 | 68.7 | 63.9 | 66.2 | 89.1 | 82.2 | 85.5 | 76.1 | 71.3 | 73.6 |
| Peregrine | 63.5 | 64.3 | 63.9 | 56.6 | 57.3 | 56.9 | 77.1 | 78.4 | 77.7 | 64.6 | 65.3 | 64.9 |
| Peregrine+NLP | 82.2 | 74.2 | 78.0 | 73.5 | 66.4 | 69.8 | 89.6 | 81.6 | 85.4 | 76.7 | 70.2 | 73.3 |
AZDC, Arizona Disease Corpus; NLP, natural language processing.